From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (qmail 9658 invoked by alias); 24 Dec 2009 16:35:04 -0000 Received: (qmail 9650 invoked by uid 22791); 24 Dec 2009 16:35:03 -0000 X-SWARE-Spam-Status: No, hits=-1.8 required=5.0 tests=AWL,BAYES_00,KAM_STOCKGEN,SPF_PASS X-Spam-Check-By: sourceware.org Received: from mail.codesourcery.com (HELO mail.codesourcery.com) (38.113.113.100) by sourceware.org (qpsmtpd/0.43rc1) with ESMTP; Thu, 24 Dec 2009 16:35:00 +0000 Received: (qmail 16092 invoked from network); 24 Dec 2009 16:34:58 -0000 Received: from unknown (HELO orlando.local) (pedro@127.0.0.2) by mail.codesourcery.com with ESMTPA; 24 Dec 2009 16:34:58 -0000 From: Pedro Alves To: gdb-patches@sourceware.org Subject: Re: [PATCH] Handle OP_THIS in tracepoints Date: Thu, 24 Dec 2009 16:35:00 -0000 User-Agent: KMail/1.9.10 Cc: Stan Shebs References: <4B32C4ED.6050904@codesourcery.com> In-Reply-To: <4B32C4ED.6050904@codesourcery.com> MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Content-Disposition: inline Message-Id: <200912241634.55084.pedro@codesourcery.com> X-IsSubscribed: yes Mailing-List: contact gdb-patches-help@sourceware.org; run by ezmlm Precedence: bulk List-Id: List-Subscribe: List-Archive: List-Post: List-Help: , Sender: gdb-patches-owner@sourceware.org X-SW-Source: 2009-12/txt/msg00381.txt.bz2 On Thursday 24 December 2009 01:33:33, Stan Shebs wrote: > This one comes up pretty quickly when doing tracepoints in C++; members > in expressions use OP_THIS, which amounts to shorthand for a reference > to the implicit argument "this". Handling is uncomplicated, mimics > expression evaluation. > > + case OP_THIS: > + { > + char *name; > + struct frame_info *frame; > + struct symbol *func, *sym; > + struct block *b; > + > + name = current_language->la_name_of_this; > + if (!name) > + error (_("no `this' in current language")); > + > + frame = get_selected_frame (_("no frame selected")); > + > + func = get_frame_function (frame); > + if (!func) > + error (_("no `%s' in nameless context"), name); > + > + b = SYMBOL_BLOCK_VALUE (func); > + if (dict_empty (BLOCK_DICT (b))) > + error (_("no args, no `%s' in block"), name); > + > + /* Calling lookup_block_symbol is necessary to get the LOC_REGISTER > + symbol instead of the LOC_ARG one (if both exist). */ > + sym = lookup_block_symbol (b, name, NULL, VAR_DOMAIN); > + if (!sym) > + error (_("no `%s' found"), name); > + > + gen_var_ref (exp->gdbarch, ax, value, sym); > + (*pc) += 2; > + } > + break; > + > case OP_TYPE: > error (_("Attempt to use a type name as an expression.")); > This is busted, and was later fixed in our internal tree. :-) The selected frame, and the current language don't have anything to do with the context in which the tracepoint runs. E.g.: (gdb) trace 'Foo::Bar()' (gdb) actions (gdb) collect foo_field Here, the `this' for foo_field needs to be looked up in the context of Foo::Bar, not of whatever context/frame the user had selected when she issued the 'actions' command. (Note that `maint agent(-eval)' always works in the context of the selected frame, so testing with that alone can mask out these issues.) -- Pedro Alves